How to read an animal study

A headline that says a treatment worked in mice is a statement about mice. It is a reason to test the treatment in people, not evidence that it works in people. This guide sets out what an animal result can and cannot show, and how to check the report itself.

This guide is general information, not medical advice.

What the report should contain

The ARRIVE guidelines describe themselves as "a checklist of recommendations for the full and transparent reporting of research involving animals", and they aim at "maximising the quality and reliability of published research" (https://arriveguidelines.org/).

The ARRIVE guidelines page (https://arriveguidelines.org/arrive-guidelines) lists ten Essential items, which it describes as the basic minimum for any manuscript describing animal research: study design, sample size, inclusion and exclusion criteria, randomisation, blinding or masking, outcome measures, statistical methods, experimental animals, experimental procedures and results.

The same page lists a Recommended Set. Item 17 covers interpretation and scientific implications, and item 18 covers generalisability and translation. Items 19 and 21 cover protocol registration and declaration of interests.

What an animal result can and cannot show

A 2013 paper in the Proceedings of the National Academy of Sciences, "Genomic responses in mouse models poorly mimic human inflammatory diseases" (Seok J and colleagues, DOI 10.1073/pnas.1222878110, first published 11 February 2013), compared genomic responses to acute inflammatory stresses in humans and in mouse models. Its abstract says that the human responses were highly similar across different causes, but that the mouse responses correlated poorly with the human conditions and with each other (https://pmc.ncbi.nlm.nih.gov/articles/PMC3587220/). That is one comparison in one set of conditions. It is still a reason to test a mouse result in people rather than assume that it carries over.

Four checks a reader can make

1. Sample size. Is the number of animals per group given, and is there a reason for it? Sample size is one of the Essential items (https://arriveguidelines.org/arrive-guidelines).
2. Randomisation and blinding. Does the report say how animals were allocated to groups, and whether the people assessing outcomes knew which group was which? Randomisation and blinding or masking are both Essential items (https://arriveguidelines.org/arrive-guidelines).
3. Species, strain and sex. Does the report name the animals in full? "Experimental animals" is an Essential item (https://arriveguidelines.org/arrive-guidelines). A result in one strain or one sex may not hold in the others.
4. Translation. Does the discussion say what the result means for people, and does it state its limits? Generalisability and translation is item 18 in the Recommended Set (https://arriveguidelines.org/arrive-guidelines). A discussion that moves from animals to patients without this step makes a claim that the study design does not support.

Reading the headline

When a news story reports an animal result, find the original report and check the four points above before repeating the claim. If the report does not contain the Essential items, say so. Do not rely on the press release for those details.

What this guide does not cover

We do not give a figure for how often animal results later hold up in people, because the sources used in this guide do not give one.